I am experiencing a recurring and intermittent failure with the Code Interpreter in ChatGPT during continuous and extended sessions. Once used, the Code Interpreter fails to reactivate after a period of inactivity, and returns incorrect results for simple operations. Despite this issue, there are scenarios where the Code Interpreter operates without problems, which I highlight to better understand the conditions under which the tool operates correctly.
Failure Scenarios:
- Activate the Code Interpreter to perform an operation, such as adding 1+1.
- Leave the session inactive for a few hours or until the next day.
- Attempt to reuse the Code Interpreter in the same conversation, resulting in access failure or erroneous responses, such as the sum of 1+1 resulting in 6.2.
Success Scenarios:
- Initial Activation and Immediate Use: In a new or existing chat session, the Code Interpreter can be successfully activated for the first time, executing any code without errors, provided it is the first activation in that conversation.
- Continuity Without Interruption: If the Code Interpreter is continuously used without long periods of inactivity, it continues to function normally, processing multiple code requests.
- Session Restarts: Even in conversations that were started weeks ago, as long as the Code Interpreter had not been used previously, it can be activated without problems the first time when the conversation is resumed.
Expected Result: The Code Interpreter should maintain consistency, allowing multiple activations in a session, with the ability to reactivate after periods of inactivity, and always returning correct results.
Impact: The intermittent failure limits the usefulness of the Code Interpreter for extended work sessions and affects the reliability of the tool in scenarios that require flexibility and resilience.
